Re: [RFA/RFC] extract strings from m4 skeletons

[Tim Van Holder]

Paolo Bonzini wrote:
> In the other case, instead of saying "%define variable `foo'" we have to
> say `%define foo'" to make the translator's life easier.  The
> problematic case occurred in bison.m4, but I did the same change to the
> wording in parse-gram.y too, for consistency.

Is there no way to expand macros or whatever during your extraction, so
that the nicer "%%code qualifier `%s' not used" message ends up in the
strings file?  I'm slightly concerned we're losing the extra 'variable'
and 'qualifier' words from messages here, which may make them less
clear.

> Anyway, back to the meat of the patch, which is data/m4-xgettext.sh.
> 
> 1) What does it do?  The scripts takes the list of skeletons and the
> list of macros to trace, and produces a `fake' C output with the
> strings.  This C output is then fed to xgettext.  Producing C has the
> advantage that xgettext understands #line directives: this way the m4
> source file names end up in bison.pot, instead of the fake C file.

I suppose this is the best way to handle it - it prevents us from
having to worry about correct .po syntax.

> 3) Is it good enough?

>From a translator's viewpoint, as long as it doesn't make the
localizable strings harder to grok (and as stated above, I'm slightly
concerned that this is the case), it's good enough.